Journal of Counseling Psychology
1998, Vol. 45, No. 1,79-83
Copyright 1998 by the American Psychological Association, Inc.
0022-0167/98/$3.00
Internal Working Models, Self-Reported Problems, and Help-Seeking
Attitudes Among College Students
Frederick G. Lopez, Mickey C. Melendez, Eric M. Sauer, Ellie Berger, and James Wyssmann
Michigan State University
A person's internal working models of close relationships (J. Bowlby, 1988) incorporate 2
discrete yet interrelated cognitive schemas: a self model containing perceptions of one's own
worth and lovability and an other model embodying core expectations about the trustworthiness and dependability of intimate others in one's social world. This study tested hypothesized
interrelationships, in a college-student sample, (a) between self-model differences and
self-reported problems and (b) among other-model differences, problem levels, and helpseeking attitudes. As expected, students with positive self models reported significantly fewer
problems than did students with negative self models. In addition, students' other-model
grouping significantly interacted with their problem levels to predict their expressed
willingess to seek counseling. Implications of the findings for counseling practice are
discussed.
The acknowledgement of personal problems does not, in
and of itself, propel affected persons to seek therapeutic
help. Presumably, those persons who voluntarily pursue
counseling must not only be experiencing distress but must
also be inclined, under these circumstances, to perceive
others as potential sources of help and support. Because
those individual differences that dispose persons to experience problems may be distinct from those that orient them
favorably toward seeking help from others, it is important to
consider how these person factors may be conceptually
interrelated.
In this study, we used contemporary attachment theory
(Bowlby, 1988) to generate and test hypotheses regarding
these individual differences. In particular, we focused on
Bowlby's concept of internal working models of attachment
as a means of operationalizing (a) subgroups assumed to
differ in their dispositions to report personal problems and
(b) other subgroups assumed to differ in their inclinations to
seek therapeutic help when experiencing high levels of
problems.
relationships (Kobak & Sceery, 1988). A basic assumption
of the theory is that critical variations in the quality of one's
early experiences with caregiving figures—particularly
around themes of separation, distress, and reunion—shape
the formation of one's internal working models of close
relationships. These models incorporate two discrete yet
interrelated cognitive schemas: a self model containing basic
perceptions of one's own worth, competence, and lovability
and an other model embodying core expectations regarding
the essential goodness, trustworthiness, and dependability of
important others in one's social world. Bowlby (1988)
conjectured that a person's internal working models, once
formed in early childhood, serve as a relatively enduring
template for his or her subsequent intimate (adult) relationships. He further argued that these attachment-related schemas are likely triggered when one is stressed, fatigued, or
ill—internal states that activate the attachment system and,
depending on the configuration of one's internal working
models, determine whether one pursues proximity seeking
or proximity avoidance.
More recently, Bartholomew and Horowitz (1991) proposed a taxonomy of adult attachment styles premised on the
relative valences (positive vs. negative) of persons' internal
working models of self and other. According to their
taxonomy, individuals with secure and dismissive attachment styles are each presumed to have positive self models
and thus are likely to view themselves as competent and
worthy of love; however, dismissive individuals, unlike their
secure counterparts, have internalized negative models of
others, which lead them to deny or discount the importance
of close relationships and instead to maintain a posture of
rigid self-sufficiency and counterdependence. Alternatively,
persons with preoccupied and fearful attachment styles are
both assumed to have internalized models of self that lead
them to doubt their own competence and efficacy; yet
preoccupied persons, given their positive models of others,
are presumed to seek proximity to others when distressed,
whereas fearful persons, given their negative models of
Internal Working Models of Self and Other
Attachment theory has been described as a framework for
understanding affect regulation in the context of close
Frederick G. Lopez, Mickey C. Melendez, Eric M. Sauer, Ellie
Berger, and James Wyssmann, Department of Counseling, Educational Psychology, and Special Education, Michigan State University. The contributions of the four junior authors were equal, and
their order of authorship was randomly determined.
We thank James Fuendeling, Kim Thomas, and David Sagula for
their assistance with data collection. We also thank Anne Mauricio,
Tracy Smith, and Ken Rice for their comments on earlier versions
of the article.
Correspondence concerning this article should be addressed to
Frederick G. Lopez, Department of Counseling, Educational
Psychology, and Special Education, 441 Erickson Hall, Michigan
State University, East Lansing, Michigan 48824.
79
80
LOPEZ, MELENDEZ, SAUER, BERGER, AND WYSSMANN
others, would be less inclined to do so. In short, the
self-model and other-model components of one's internalized attachment schemas permit differential predictions
regarding one's (a) disposition to experience and report
problems and (b) inclinations to seek or to avoid help when
distressed.
engage in less self-disclosure, and make poorer use of
treatment. More recently, Satterfield and Lyddon (1995)
observed that clients expressing attachment-related distrust
in the availability and dependability of others (i.e., negative
models of others) were more likely to evaluate the early
phase of the counseling relationship in negative terms.
Working Models, Self-Reported Problems,
and Help Seeking
The Present Study
Empirical support for the relative independence of the
self- and other-model components of attachment schemas,
as well as for their differential links to distress and to help
seeking, comes from several sources. In a series of three
studies that used self-reports, friend-reports, and trained
judges' ratings of participants' attachment orientations,
Griffin and Bartholomew (1994) demonstrated the construct
validity of the self- and other-model dimensions of adult
attachment. Whereas self-model and other-model intercorrelations were statistically insignificant, self models showed
strong associations with self-concept/self-esteem indicators,
whereas other models demonstrated equally strong correspondences with measures of interpersonal orientation. Griffin
and Bartholomew concluded that symptom reporting would
be expected to "line up with the self-model dimension" (p.
443).
Consistent with this speculation is the finding by Camelley, Pietromonaco, and Jaffe (1994) that mildly depressed
college women, relative to their nondepressed peers, held
negative models of self but both positive and negative
models of others. Elsewhere it was shown that, relative to
their secure and dismissive peers, persons with negative self
models (i.e., preoccupied and fearful attachment styles)
demonstrated greater shame proneness and deficient problem solving in relationships (Lopez et al., 1997) and
experienced greater emotional distress following relationship terminations (Pistole, 1995; Simpson, 1990). By contrast, Kobak and Sceery (1988) found that dismissive
persons, despite obtaining peer ratings of lower egoresilience, higher anxiety, and greater hostility, did not differ
from their secure counterparts on self-report measures of
perceived social competence and distress. Kobak and Sceery
speculated that this lack of congruence between peer reports
and self-reports may reflect a tendency among dismissive
individuals to deny the experience of negative affect.
To date, few empirical studies have directly extended
attachment theory to the study of counseling processes
(Mallinckrodt, Gantt, & Coble, 1995); thus, evidence that
other-model differences are predictive of therapeutic helpseeking attitudes is largely indirect. For example, persons
with negative models of others have been shown to exhibit
the least comfort and competence with self-disclosure
(Bartholomew & Horowitz, 1991; Mikulincer & Nachson,
1991; Pistole, 1993), which suggests a likely orientation
away from therapeutic help seeking. Consonant with this
speculation is Dozier's (1990) finding that among a group of
clients with serious psychopathological disorders, dismissive adults were more likely than their counterparts with
other attachment orientations to reject treatment providers,
Taken together, the above findings suggest that whereas
participants' self models should predict the level of their
self-reported problems, their other models should predict
their therapeutic help-seeking attitudes. Although a few
studies have begun to examine adult attachment differences
within samples already receiving therapeutic services, we
could locate no studies that concurrently examined indexes
of distress and help-seeking attitudes within a nonclinical
sample.
To pursue this inquiry, in the present study we advanced
and tested the following hypotheses. First, we anticipated
that persons with positive self models (i.e., those reporting
either secure or dismissive attachment styles) would report
significantly fewer personal problems than would their peers
with negative self models (i.e., preoccupied and fearful
styles). Second, we hypothesized that participants' other
models and their current problem levels would interact to
predict their help-seeking attitudes. More specifically, we
expected that among persons reporting high levels of
problems, those individuals with positive models of others
(i.e., "secures" and "preoccupieds") would acknowledge
more favorable orientations toward therapeutic help seeking
than would their peers with negative models of others (i.e.,
"dismissives" and "fearfuls").
Method
Participants
Two-hundred fifty-three college students (95 men, 157 women,
and 1 person of undisclosed gender) were recruited from undergraduate education and psychology classes at a large midwestern
university to participate in a study of "factors affecting college
student distress and help-seeking preferences." Participants were
predominantly either juniors (27%) or seniors (34%) who were
single and had never been married (96%). The mean age of the
sample was 20.95 years (SD = 4.25; range = 17-48), and ethnic
group representations were as follows: White (78%); African
American (14%); Asian American (2%); Hispanic-Latino/a (2%);
multiracial (2%); and undeclared (2%). One hundred eighty-two
participants (72%) indicated that they had never sought counseling
for a personal problem, whereas the remainder of the sample
(/i = 71, 28%) had.
Measures
Demographic questionnaire. This brief questionnaire gathered
background information on students' age, gender, year in college,
marital status, and current relationship or dating status. Respondents were also asked to indicate whether they had ever sought and
received counseling for a personal problem.
Self-reported problems. A 20-item version of the Personal
Problems Inventory (PPI; Cash, Begley, McCown, & Weise, 1975)
81
INTERNAL WORKING MODELS
that was modified by Ponce and Atkinson (1989) was used to assess
participants' current levels of self-reported problems. Participants
rated each of 20 problems commonly experienced by college
students according to its current severity in their lives using a
6-point rating scale (1 = not at all a problem; ; 6 = verysignificant problem). Ponce and Atkinson added five items to the
original 15-item measure; these items assessed concerns regarding
adjustment to college, academic performance, financial problems,
and feelings of loneliness, isolation, and not belonging. The authors
also reported that two moderately intercorrelated (r = .65) factors
best described PPI item variation, with personal-social problems
(i.e., loneliness-isolation, dating difficulties) loading heavily on the
first factor and academic adjustment problems (i.e., academic
performance, career choice) loading on the second factor. For the
purposes of the present study, all 20 PPI items were summed to
produce a total problem score, which yielded an internal consistency (Cronbach's alpha) coefficient of .83. Those participants in
our sample who acknowledged receiving prior counseling reported
significantly higher PPT total scores than did their peers who had
not, FU, 249) = 9.31,p<.01.
Willingness to seek counseling. Following their ratings of
current problems, participants were asked, in a subsequent series of
ratings, to indicate their willingness to see a counselor for help with
each of the PPI problem items, regardless of whether it was
currently a problem in their lives. Respondents indicated their
willingness to seek counseling for each of the 20 problem items
using a 6-point rating scale (1 = not at all willing; 6 = very
willing), and item ratings were summed to produce a total
willingness score. Solberg, Ritsma, Davis, Tata, and Jolly (1994)
reported a Cronbach's alpha of .94 for their PPI-derived willingness scale; they also found that willingness scores were positively
correlated with their participants' prior counseling experience. In
the present study, the obtained Cronbach's alpha coefficient for the
willingness scale was .95. Willingness scores were also unrelated
to PPI problem scores (see Table 1).
Attitudes toward seeking professional help. Our participants'
help-seeking attitudes were additionally assessed by their responses to the Attitudes Toward Seeking Professional Psychological Help Scale (ATSPPHS; Fischer & Turner, 1970). Developed on
a large, college-student sample, this 29-item questionnaire taps
attitudes indicative of a general orientation toward seeking therapeutic help for psychological problems (e.g., "Emotional difficulties,
like many things, tend to work out by themselves" [reverse-scored]
and "At some future time, I might want to have psychological
counseling"). Respondents indicate their level of agreement with
each item on a 4-point scale, and following appropriate item
receding, item ratings are summed, with higher scores indicating
more favorable attitudes. Fischer and Turner identified four factors
underlying interitem variation: (a) Need (i.e., perceived need for
Table 1
professional help), (b) Stigma (i.e., tolerance of stigma associated
with receiving counseling services), (c) Openness (i.e., interpersonal openness with regard to reflecting upon and revealing one's
personal problems), and (d) Confidence (i.e., confidence in mental
health professionals to be of practical assistance). Items across
subscales can also be summed to produce a total help-seeking
score. Because of the modest internal consistencies of some
factor-derived subscales, Fischer and Turner recommended use of
the overall ATSPPHS score (Cronbach's alpha = .83). In the
present study, the obtained Cronbach's alpha coefficient for the
ATSPPHS total scale was .89.
Fischer and Turner (1970) reported (a) significant gender
differences on their measure, with women endorsing more favorable help-seeking attitudes than men, (b) that ATSPPHS total
scores reliably distinguished persons who had previously obtained
psychotherapeutic help from those who had not, and (c) that these
scores evidenced considerable test-retest stability over periods
ranging from 5 days (r — .86) to 2 months (r = .84).
Relationship Questionnaire (RQ; Bartholomew & Horowitz,
1991). This self-classification measure of adult attachment style
asks respondents to indicate which of four descriptive paragraphs
best describes how they feel about close relationships. The four
paragraphs respectively characterize secure, dismissive, preoccupied, and fearful attachment styles. Bartholomew (1989) reported
that the four RQ self-classifications demonstrated moderate stability ratings over a 2-month period: secure, .71; dismissive, .49;
preoccupied, .59; and fearful, .64. In addition, Bartholomew and
Horowitz (1991) found with a college sample that the four
attachment styles related in theoretically consistent ways with both
self-reports and friend-reports of respondents' self-esteem and
sociability. More recently, Scharfe and Bartholomew (1994) reported that 63% of the women and 56% of the men in their young
adult sample retained the same RQ self-classification over an
8-month interval. Self-ratings indicating a secure attachment style
were especially stable, with 71% of women and 61% of men
retaining their secure self-classification across this time period.
Indeed, RQ stability coefficients were comparable to those obtained by both interview-based and continuously scaled self-report
measures of adult attachment style. Elsewhere, Kirkpatrick and
Hazan (1994) reported an overall 70% correspondence rate in
attachment style self-classification over a 4-year period within their
adult sample.
Procedure
Recruitment solicitations were briefly conducted at the beginning of regular class meetings. Interested students attended one of
several scheduled group testing sessions, during which time they
read and signed informed consent forms and then completed a
survey packet containing the measures described above. All
respondents received some course credit for their participation.
Results
Intercorrelations Among Problem and
Help-Seeking Measures
Measure
1
1. PPI total score
2. ATSPPHS total score
3. Willingness to seek counseling
—
.01
.10
.02
—
.31*
.09
.42*
—
Note. N - 250. Scale intercorrelations among men are above the
diagonal; intercorrelations among women are below the diagonal.
PPI - Personal Problem Inventory; ATSPPHS = Attitudes Toward
Seeking Professional Psychological Help Scale.
0
In view of our mixed-gender sample and prior studies
indicating gender effects on problem self-reports (Kenny &
Donaldson, 1991; Wohlgemuth & Betz, 1991), we tested our
first hypothesis using a 2 X 2 (Gender X Self-Model)
analysis of variance (ANOVA) of PPI total scores. Results
indicated a highly significant effect for self-model group,
F(l, 245) = 38.33, p < .001, with participants in the
positive self-model group (i.e., those with secure and
dismissive attachment styles) acknowledging significantly
fewer problems than did their peers in the negative self-
82
LOPEZ, MELENDEZ, SAUER, BERGER, AND WYSSMANN
model group (i.e., participants with preoccupied and fearful
styles). No significant effects were observed either for
participants' gender or for its interaction with self-model
grouping.
Prior to testing our hypothesis regarding help-seeking
attitudes, we used a median split of PPI total scores in our
sample to classify our participants into low and high
problem-level groups. In light of the fact that our two
help-seeking measures were only moderately intercorrelated, and recalling that significant gender differences in
help-seeking attitudes had been previously observed (Fischer
& Turner, 1970), we then used a 2 X 2 X 2 ANOVA
(Gender X Problem Level X Other Model) on each index,
adjusting alpha for the two tests (.05/2, p = .025).
Results indicated that, with regard to ATSPHS total
scores, only participants' gender yielded a significant effect,
F(l, 241) = 70.20, p < .001, with women reporting a more
favorable general orientation to help seeking than did men.
No other main or interaction effects were observed on this
measure. Significant gender differences were similarly observed on our index of willingness to seek counseling, F(l,
241) = 16.30,/? < .001; in addition, however, the expected
interaction of problem level and other-model grouping was
observed on this measure, F{\, 241) = 5.18,/? < .025. This
final analysis yielded no other significant main or interaction
effects. Figure 1 illustrates the nature of the significant
interaction of problem level and other-model grouping on
participants' willingness to seek counseling. In line with our
hypothesis, among participants acknowledging high current
problem levels, those with positive other models (i.e., those
with secure and preoccupied attachment styles) demonstrated discernibly higher scores than did those with nega-
80
POSITIVE OTHER
MODEL
"5
oI
70
CO
I 60
NEGATIVE OTHER
MODEL
50
LOW
HIGH
Problem Level
Figure 1. Interaction of other-model group and problem level on
willingness to seek counseling.
tive other models (i.e., those with dismissive and fearful
attachment styles).
Discussion
As anticipated, our participants' self models significantly
predicted their overall level of self-reported problems:
Those individuals with positive self models acknowledged
significantly fewer current problems than did their counterparts with negative self models. These results are in line with
previous reports (Bartholomew & Horowitz, 1991; Horowitz, Rosenberg, & Bartholomew, 1993) (a) indicating that
students with negative self models (i.e., preoccupied and
fearful attachment styles) are more inclined to experience
adjustment difficulties and (b) speculating that symptom
reporting would be expected to "line up with the self-model
dimension" (Griffin & Bartholomew, 1994, p. 443). Moreover, although we did not specifically test for it in our
primary analyses, a post hoc comparison indicated there was
no significant difference in the PPI means of the two
attachment styles representing the positive self-model group
(i.e., secures and dismissives), which thus lends further
support for this view. It is important to note, however, that
dismissive persons have demonstrated elevated activity on
physiological measures while simultaneously denying the
experience of distress (Dozier & Kobak, 1992), a finding
suggesting that subtle differences in the quality of the
positive self models internalized by persons with secure and
dismissive attachment styles may indeed predict problem
scores on measures less vulnerable to self-report bias.
Our hypothesis that participants' overall problem levels
would significantly interact with the relative valences of
their other models to predict their help-seeking attitudes was
supported with regard to one of our help-seeking measures
(willingness to seek counseling) but not for the other
(ATSPPHS scores). Because these measures were only
moderately correlated within our sample, this inconsistent
finding may be a function of important differences between
these measures. As noted earlier, ATSPPHS items assessed a
general orientation to therapeutic help seeking; the willingness measure, by contrast, more specifically directed respondents to indicate their own orientation toward pursuing
counseling should problems arise in their lives. Perhaps the
more personal focus of the latter measure was necessary,
along with a heightened level of self-reported problems, to
activate respondents' core expectations of the potential
trustworthiness and dependability of others.
Our finding that persons with high levels of problems and
negative models of others are less willing to seek counseling
may, in part, explain why such individuals who eventually
reach the counselor's office evaluate the early phase of the
counseling relationship in negative terms (Satterfield &
Lyddon, 1995) and may be prone to prematurely terminate
therapeutic relationships (Mallinckrodt et al., 1995). Because a thorough assessment of clients' internal working
models may require sufficient time, counselors are well
advised to explore (from the outset of counseling and in at
least a preliminary fashion) whether their clients are ambivalent about seeking help for their problems (Teyber, 1997).
83
INTERNAL WORKING MODELS
Such early discussions may be useful in identifying (and
later addressing) potentially problematic expectancies regarding the counselor's competence, trustworthiness, and dependability that may be part of their clients' more general,
internalized working models of others.
Although our findings clarify interrelationships among
college students' internal working models, self-reported
problems, and help-seeking attitudes, some important limitations of this study should be noted. First, our use of a field
correlational design does not permit cause-effect conclusions to be drawn about the relations of working models to
problem reports or to help-seeking attitudes. Second, we
relied exclusively on self-report instruments to measure our
key constructs. As noted earlier, persons with dismissive
attachment styles are inclined to deny negative affect and
personal vulnerability, which may lead them to underreport
the experience of personal problems in their lives. Therefore, future studies would do well to combine self-report,
informant-report, physiological, and behavioral problem
measures, as well as interview-based and observational
methodologies, in order to more sensitively assess the key
constructs in this investigation. Third, the fact that all of our
participants were drawn from undergraduate education and
psychology courses may have introduced an unknown
selection bias. Replication of this study both within a more
diverse sample as well as among individuals concurrently
receiving counseling services is therefore advised.
References
Bartholomew, K. (1989). [Stability of four-category attachment
patterns in young adults over two months]. Unpublished raw
data.
Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles
among young adults: A test of a four-category model. Journal of
Personality and Social Psychology, 61, 226-244.
Bowlby, J. (1988). A secure base: Parent-child attachment and
healthy human development. New York: Basic Books.
Camelley, K. B., Pietromonaco, P. R., & Jaffe, K, (1994). Depression, working models of others, and relationship functioning.
Journal of Personality and Social Psychology, 66, 127-140.
Cash, T. R, Begley, P. J., McCown, D. A., & Weise, B, C. (1975).
When counselors are heard but not seen: Initial impact of
physical attractiveness. Journal of Counseling Psychology, 22,
273-279.
Dozier, M. (1990). Attachment organization and treatment use for
adults with serious psychopathological disorders. Development
and Psychopathology, 2, 47-60.
Dozier, M., & Kobak, R. R. (1992). Psychophysiology in attachment interviews: Converging evidence for deactivating strategies. Child Development, 63, 1473-1480.
Fischer, E. H., & Turner, J. L. (1970). Orientations to seeking
professional help: Development and research utility of an
attitude scale. Journal of Consulting and Clinical Psychology,
35, 79-90.
Griffin, D., & Bartholomew, K. (1994). Models of the self and
other: Fundamental dimensions underlying measures of adult
attachment. Journal of Personality and Social Psychology, 67,
430-445.
Horowitz, L. M., Rosenberg, S. E., & Bartholomew, K. (1993).
Interpersonal problems, attachment styles, and outcome in brief
dynamic psychotherapy. Journal of Consulting and Clinical
Psychology, 61, 549-560.
Kenny, M. E., & Donaldson, G. A. (1991). Contributions of
parental attachment and family structure to the social and
psychological functioning of first-year college students. Journal
of Counseling Psychology, 38, 479-486.
Kirkpatrick, L. A., & Hazan, C. (1994). Attachment styles and
close relationships: A four-year prospective study. Personal
Relationships, 1, 123-142.
Kobak, R. R., & Sceery, A. (1988). Attachment in late adolescence:
Working models, affect regulation, and representations of self
and others. Child Development, 59, 135-146.
Lopez, F. G., Gover, M. R., Leskela, J., Sauer, E. M., Schirmer, L.,
& Wyssmann, J. (1997). Attachment styles, shame, guilt, and
collaborative problem-solving orientations. Personal Relationships, 4, 187-199.
Mallinckrodt, B., Gantt, D. L., & Coble, H. L. (1995). Attachment
patterns in the psychotherapy relationship: Development of the
Client Attachment to Therapist Scale. Journal of Counseling
Psychology, 42, 307-317.
Mikulincer, M., & Nachson, O. (1991). Attachment styles and
patterns of self-disclosure. Journal of Personality and Social
Psychology, 61, 321-331.
Pistole, M. C. (1993). Attachment relationships: Self-disclosure
and trust. Journal of Mental Health Counseling, 15, 94-106.
Pistole, M. C. (1995). College students' ended love relationships:
Attachment style and emotion. Journal of College Student
Development, 36, 53-60.
Ponce, F. Q., & Atkinson, D. R. (1989). Mexican-American
acculturation, counselor ethnicity, counseling style, and perceived counselor credibility. Journal of Counseling Psychology,
36, 203-208.
Satterfield, W. A., & Lyddon, W. J. (1995). Client attachment and
perceptions of the working alliance with counselor trainees.
Journal of Counseling Psychology, 42, 187-189.
Scharfe, E.T & Bartholomew, K. (1994). Reliability and stability of
adult attachment patterns. Personal Relationships, 1, 23-43.
Simpson, J. A. (1990). Influence of attachment styles on romantic
relationships. Journal of Personality and Social Psychology, 59,
971-980.
Solberg, V. S., Ritsma, S., Davis, B. J., Tata, S. P., & Jolly, A.
(1994). Asian-American students' severity of problems and
willingness to seek help from university counseling centers:
Role of previous counseling experience, gender, and ethnicity.
Journal of Counseling Psychology, 41, 275-279.
Teyber, E. (1997). Interpersonal process in psychotherapy (3rd
ed.). Pacific Grove, CA: Brooks/Cole.
Wohlgemuth, E., & Betz, N. E. (1991). Gender as a moderator of
the relationships of stress and social support to physical health in
college students. Journal of Counseling Psychology, 38, 367374.
Received February 10, 1997
Revision received August 4,1997
Accepted August 6,1997